[00:01] morning brief is ready. [music] Three new leads came in and your client report is drafted. >> That Jarvis voice was made for free on my Mac in an app called Voice Studio in about 20 seconds. No cloud and nothing [00:16] leaves my computer. Today I'll show you how to build your own Jarvis style voice with it, plus a whole cast of voices for your business. Listen to these. An agency [music] sending a client update. Hi Sarah, quick update from the team. [00:30] Your new landing page is live and you moved up on four keywords this week. A coach with a daily check-in. Morning. It's day 12. Before you open your inbox, take 2 minutes and write down your one big win for today. A course owner [00:46] opening a lesson. >> Welcome to module 3. In this lesson, you'll build your first AI agent step by step in under 10 minutes. A software [00:58] founder onboarding a new user. >> Welcome aboard. Click the blue button to connect your inbox and I'll set up your first workflow for you. first workflow for you. >> And a creator opening a video. [01:12] >> Want your AI assistant to have a real voice? Stick around because by the end of this video, yours will. >> Every one of those came out of the same free app running offline on a normal Mac. By the end of this video, you'll [01:26] have the exact voice prompts, the setup, and five ways to put these voices to work, including giving your own Jarvis a real voice. So, what is Voice Studio? It's a free open-source voice studio that runs on your own computer. It's on [01:41] that runs on your own computer. It's on GitHub. It has over 111,000 stars, and the latest release came out on the 12th of November. It does text to speech, voice design, voice cloning, dubbing, transcription, and long- form audio, all [01:54] locally. Voice design is the part I love. You don't need a recording of anyone. You type a description of the voice you want, and it builds that voice from scratch. And here's why this matters for you. A member in the AI [02:07] profit boardroom asked this week, who's actually running AI offline on a Mac and how to set it up without a list of 20 tools. This is one of the easiest wins for that. One app, one install, and your voices live on your own machine. Let me [02:22] show you the install. On a Mac, you open terminal and paste three homebrew commands. You add the Voice Studio tap. You trust the cask and you install it. If you're on Windows or Linux, there's an installer on the GitHub releases page [02:35] instead. The first time you open it, Mac OS asks if you're sure because it was downloaded from the internet. Click open. Then, Voice Studio sets up its local runtime. That's Python and the AI libraries. And it shows you how much [02:49] disk space you need before it starts. Mine took a few minutes. Next, it runs a system check. Operating system, memory, disk space, and whether your Apple chip can run the models on the graphics cores. Everything ticked green on my 16 [03:03] GB Mac. Then you pick a model pack. I chose balanced. It downloads the voice model and the transcription models together. Quick one before we build the Jarvis voice. If you want my full voice studio setup, the exact voice prompts [03:17] from this video, and the workflows that plug these voices into your business, they're inside the AI profit boardroom. You'll also get the faceless video skill I used to make this video, daily updates on new AI tools, and a community of [03:30] business owners who share what's working for them. There are weekly coaching calls where you can bring your own setup and get help with it. The link is in the comments and the description or go to aiprofitboardroom.com. [03:42] Then there's a privacy screen. I said no to sharing stats and I left the invisible watermark switched on. That marks every file as AI made, which is the honest way to use this stuff. Now, one thing most people will miss and it's [03:56] important. The default voice model in Voice Studio is licensed for personal and research use only. The app actually tells you this in settings under models. Every engine has a license badge and the default one is red. It says commercial [04:12] use of the generated audio is not allowed. So if you're using these voices for your business, switch to an engine with a green badge. My first pick was Vox CPM2. It's released under Apache 2.0, so you can use the audio [04:26] commercially and it does voice design, but on my 16 GB Mac, it was too heavy. and one line took minutes. So, I opened MLX audio, the engine pack built for Apple chips, and downloaded Quen 3TS Voice Design. It's also Apache 2.0. It's [04:44] only 1.4 GB and it builds voices from a description. Got a big machine? Vox CPM2 is a great choice. On a normal Mac, use Quen 3. Right, let's build Jarvis. You [04:56] open the voice tab, then design. At the top there's a box that says describe your voice. This is where the magic happens. Here's the exact prompt I typed. A calm, refined British male AI butler in his 40s. Smooth, warm [05:12] baritone, crisp and precise, quietly witty, unhurried, close mic studio quality. Then you paste the script you want it to read into the big box on the right and click synthesize audio. The first run takes a little longer while [05:26] the model loads. After that, each line comes back in seconds. Now, here's the trick. Every time you generate, you get a slightly different take. If you like the voice, but want a variation, click new seed and run it again. I made four [05:42] takes and kept the one that sounded most like a calm assistant reading my morning brief. When you find the one, type a name. I called mine Jay, and click save as voice profile. Now Jay sits in your saved voices, and you can use him for [05:56] anything. You can also start from one of the presets like narrator, corporate, or storyteller and tweak the description from there. So that's the voice. And the five business voices you heard at the start were made the exact same way, just [06:10] with a different description. Here they are so you can copy them. The agency voice, a friendly, confident female account manager in her 30s, clear and upbeat, neutral British accent, professional but warm. The coach, an [06:25] energetic, encouraging male coach in his 30s. Warm American accent, bright and motivating, natural pace. The course voice, a clear, patient female teacher in her 40s. Calm and articulate, gentle American accent. The software voice, a [06:42] crisp, modern male product guide in his 20s. Friendly and efficient, neutral American accent, clean studio sound. and the creator. A playful, high energy female host in her 20ies. Expressive and quick Australian accent sounds like [06:58] she's smiling. Notice the pattern. Who they are, their age, the tone, the accent, and how it should sound. Five parts. If a voice comes out wrong, change one part at a time so you know what made the difference. Now, let's put [07:12] these voices to work. Here are five ways, and each one comes from what members are asking about right now. Number one, client updates for agencies. Every agency owner I talk to says the same thing. Clients want to hear from [07:26] you, and written reports don't get read. So, each Friday, take your client report summary, paste it in, and send a 60-second voice update in a consistent, professional voice. Same voice every week, so it feels like a real part of [07:40] your service. And if you do SEO, you can drop in what moved this week, like new pages live and keywords that went up. Number two, daily check-ins for coaches. If you run a coaching program, the hardest part is keeping clients moving [07:55] between calls. Write 30 short check-ins once, one for each day of your program, and turn them into audio in one sitting. Then drop them into your community or your email sequence. Your clients get a voice in their ear every morning and you [08:10] didn't record a single take. Number three, lesson voiceovers for course and community owners. This one is huge. A member asked this week how to keep up with daily content without burning out. If you have written lessons, Voice [08:24] Studio can turn each one into narrated audio. It has a stories section for long- form text, so a whole module comes out in one go. Your members can listen in the car or at the gym, and you can update a lesson by editing text instead [08:39] of re-recording. Number four, demo and onboarding voiceovers for software founders. Another member asked, "How to make product videos like the polished ones you see from AI voice companies? The secret is the voice over. Record [08:53] your screen, write a short script, and give it a voice like the SAS one you heard at the start. If you change a button next month, you change one line of text and the new voice over is ready in seconds. Number five, intros and [09:07] reads for creators. If you make videos or run a podcast, use a designed voice for your intro, your channel trailer, or a recurring character. Keep the character voice saved as a profile, and every episode sounds the same. And [09:20] here's the bonus, the one that ties it all together. Give your Jarvis a voice. Voice Studio runs a local server on your Mac with an MCP connection builtin. Open integrations, click Claude Code, and copy the config it shows you into your [09:36] project. That's the whole setup. Then I asked Claude code to be my Jarvis. Read my brief file, write a short morning brief, find my saved voice called Jay, and speak it. It found Jay, wrote four sentences about my new leads, my calls, [09:51] and a report that needs review. and Voice Studio turned it into 33 seconds is. >> Good morning. Since yesterday, three new leads have come in. Bright Path Coaching [10:06] wants SEO for their course launch. Nordic Gear has product pages that aren't ranking, and Loop Studio is asking about an AI automation audit. Your discovery call with Bright Path is at 11:00 and the Acme Academy SEO report [10:22] at 11:00 and the Acme Academy SEO report needs your review before 3. The Skyline Consulting landing page has been live since Friday and Stand Up is at 3:00. The same idea works with any assistant that supports MCP. So, if you run Agent [10:38] OS or your own Javis setup, you can plug it in the same way. One tip that saves a headache, set Voice Studios MCP output mode to files so your assistant gets a file path back instead of a huge block of audio data. It's in the setup guide. [10:54] A few mistakes to avoid. First, the license. Check the badge on the engine you're using before you publish anything. Green means you're good for business use. Red means personal only. Second, only clone voices you own or [11:08] have permission to use. Use voice design for characters like Jay and keep cloning for your own voice or voices you've been given consent for. Third, keep that watermark on. And fourth, write for the ear. Short sentences, one idea each, and [11:23] read it out loud once before you generate. So, here's your plan for this week. Install Voice Studio, switch to Quen 3, paste the JPrompt, and save your first voice profile. Then pick one of the five workflows that matches your [11:38] business and ship one audio piece. That's it. If you want the exact prompts, the full setup guide, and the workflows to connect these voices to your Jarvis, they're inside the AI profit boardroom along with the faceless [11:50] video skill and the daily updates. The link is in the comments and the apiprofitboardroom.com. See you in the next one.